Hard to Be Heard: Phoneme-Level ASR Analysis of Phonologically Complex, Low-Resource Endangered Languages

TL;DR

This paper analyzes phoneme-level ASR performance on two low-resource, complex East Caucasian languages, revealing data scarcity as a key factor influencing errors and demonstrating the importance of phoneme-level evaluation.

Contribution

It introduces a phoneme-level analysis framework for low-resource languages, compares state-of-the-art models, and highlights data scarcity's impact on phoneme recognition accuracy.

Findings

Phoneme recognition accuracy correlates with training frequency.

Wav2vec2 with language-specific phoneme vocabulary performs well.

Data scarcity explains many errors attributed to phonological complexity.

Abstract

We present a phoneme-level analysis of automatic speech recognition (ASR) for two low-resourced and phonologically complex East Caucasian languages, Archi and Rutul, based on curated and standardized speech-transcript resources totaling approximately 50 minutes and 1 hour 20 minutes of audio, respectively. Existing recordings and transcriptions are consolidated and processed into a form suitable for ASR training and evaluation. We evaluate several state-of-the-art audio and audio-language models, including wav2vec2, Whisper, and Qwen2-Audio. For wav2vec2, we introduce a language-specific phoneme vocabulary with heuristic output-layer initialization, which yields consistent improvements and achieves performance comparable to or exceeding Whisper in these extremely low-resource settings. Beyond standard word and character error rates, we conduct a detailed phoneme-level error analysis. We find that phoneme recognition accuracy strongly correlates with training frequency, exhibiting a characteristic sigmoid-shaped learning curve. For Archi, this relationship partially breaks for Whisper, pointing to model-specific generalization effects beyond what is predicted by training frequency. Overall, our results indicate that many errors attributed to phonological complexity are better explained by data scarcity. These findings demonstrate the value of phoneme-level evaluation for understanding ASR behavior in low-resource, typologically complex languages.